single-cell-foundation-model-stofm

To preprocess spatial transcriptomics data for cell embedding generation, convert raw data into SToFM-ready formats like hf.dataset and data.h5ad using the official preprocessing pipeline.

Generating per-cell embeddings from spatial graphs requires running the cell encoder and SE(2) Transformer on preprocessed hf.dataset and data.h5ad files to produce actionable cell embeddings.

Yes, SToFM embeddings support downstream region segmentation and cell type annotation by providing the per-cell feature representations extracted via the SE(2) Transformer and cell encoder.

Yes, you need a local SToFM repository path containing the required data.h5ad and hf.dataset files to invoke the official preprocessing and embedding pipelines successfully.

Spatial transcriptomics data preprocessing requires inputs to be formatted as hf.dataset and data.h5ad files to properly interface with the cell encoder and SE(2) Transformer.

Spatial transcriptomics preprocessing requires the SE(2) Transformer to process spatial graphs and generate accurate per-cell embeddings for downstream region segmentation and cell type annotation.